ricardian equivalence and the efficacy of fiscal policy in
TRANSCRIPT
University of WollongongResearch Online
Faculty of Business - Economics Working Papers Faculty of Business
2009
Ricardian equivalence and the efficacy of fiscalpolicy in AustraliaShane BrittleUniversity of Wollongong
Research Online is the open access institutional repository for the University of Wollongong. For further information contact the UOW Library:[email protected]
Publication DetailsBrittle, S, Ricardian equivalence and the efficacy of fiscal policy in Australia, Working Paper 09-10, Department of Economics,University of Wollongong, 2009, 37p.
University of Wollongong Economics Working Paper Series 2008 http://www.uow.edu.au/commerce/econ/wpapers.html
RICARDIAN EQUIVALENCE AND THE EFFICACY OF FISCAL POLICY IN
AUSTRALIA
Shane Brittle School of Economics
University of Wollongong
WP 09-10 September 2009
RICARDIAN EQUIVALENCE AND THE EFFICACY OF FISCAL POLICY IN AUSTRALIA
Shane Brittle
School of Economics University of Wollongong
Northfields Avenue Wollongong NSW 2500
Australia
2
RICARDIAN EQUIVALENCE AND THE EFFICACY OF FISCAL POLICY IN AUSTRALIA
Shane Brittle
ABSTRACT
Events surrounding the global financial and economic crises of 2008 and 2009 have sparked a
renewed interest in discretionary fiscal policy. This paper considers whether private saving in
Australia behaves in a manner that is consistent with Ricardian equivalence, thus mitigating the
effects of fiscal policy, or conversely, if fiscal policy has some ability to influence real economic
activity. A model of private and public saving is estimated using the autoregressive distributed lag
approach (ARDL) to cointegration. This estimation procedure is advantageous due to its ability to
provide both short- and long-run coefficient estimates, and can accommodate coefficients for
structural breaks. Given that the Australian economy has been subject to a substantial amount of
structural change over the past 50 years, the estimations attempt to account for these structural
effects on long-run savings behaviour. Results indicate that while there is not a full Ricardian
response to changes in the fiscal stance, evidence suggests some partial offsetting behaviour –
implying that fiscal policy does elicit some (limited) impact on economic activity.
Keywords: Ricardian equivalence, fiscal policy, cointegration, structural breaks.
JEL Classifications: E21, E62, C22, H62.
3
1. Introduction
Research interest in fiscal policy waned over the 1990s, and for the most part of the 2000s, as the
“new consensus” on macroeconomic policy saw monetary policy (inflation targeting) assuming the
role of stabilising short-run fluctuations in prices and output in most advanced economies. Fiscal
policy over this period was increasingly directed toward the medium-term sustainability of
government balance sheets and allowing the automatic stabilisers to freely operate.
However, fiscal policy debates in Australia were reignited in the mid 2000s as the Howard
Government undertook a series of personal income tax cuts. At that time, the economy was
operating at or near full capacity with unemployment around 30-year lows. Critics argued that this
loosening of fiscal policy would only add to aggregate demand – leading to higher inflation and
interest rates.
Sharp falls in output associated with the global financial and economic crisis in 2008 and 2009 have
seen fiscal stimulus packages enacted in many countries, and a renewed interest in activist fiscal
policy. In a number of countries monetary policy had reached the zero bound on nominal interest
rates, leaving quantitative easing measures and fiscal policy to support aggregate demand. To
prevent a severe and prolonged global downturn, in late 2008 the International Monetary Fund
(Spilimbergo et al: 2008) called for a fiscal loosening across the advanced economies amounting to
at least 2 per cent of global gross domestic product (GDP). By mid 2009, Australia had
implemented fiscal stimulus packages amounting to around 3 per cent of GDP in 2008-09 and 2 per
cent of GDP in 2009-10 (Budget: 2009).
Considering the potential efficacy of fiscal policy, Hemming (et al: 2002) provides an excellent
survey of the international evidence on fiscal multipliers from simulations using macroeconomic
models and reduced-form specifications. In short, Hemming reports that positive fiscal shocks,
generated using estimated macroeconomic models, produce positive multipliers, with expenditure
multipliers in the range of 0.6 to 1.5 and tax multipliers in the range of 0.3 to 0.8; long-term
multipliers are generally smaller and some are negative. More recent estimates have been produced
by the Organisation for Economic Co-Operation and Development (2009), and the International
Monetary Fund (2009).
However, as Kennedy (et al: 2004) note, there is little empirical evidence on the efficacy of fiscal
policy in Australia, or estimates of fiscal multipliers. Perotti (2002) finds a positive short-term
impact spending multiplier of 0.6 for Australia, peaking at 0.8 after 14 quarters. Recent estimates
4
from the OECD (2009) suggest that the fiscal multiplier in Australia is 0.2 for tax cuts, and
increases up to 1.3 for direct government investment (such as infrastructure).
In contrast to fiscal policy having some impact on aggregate demand, Ricardian equivalence asserts
that fiscal deficits merely postpone taxes, and through the actions of altruistically motivated
individuals, budget deficits have no real affects on the economy. Barro (1974) considered the
effects on bond values and tax capitalisation of finite lives, imperfect capital markets, a government
monopoly in the production of bond ‘liquidity services’ and uncertainty about future tax
obligations. Within the context of an overlapping generations model, Barro showed that finite lives
will not be relevant for future tax liabilities so long as current generations are connected to future
generations by a chain of operative intergenerational transfers (Barro: 1974). This paper gave rise to
what is now known as the Ricardian equivalence theorem, or the Barro-Ricardo hypothesis. The key
result of Barro’s investigation being that so long as there is an operative intergenerational transfer,
there will be no net-wealth effect and no effect on aggregate demand; or on interest rates of a
marginal change in government debt. Essentially, under the Barro-Ricardo hypothesis deficits do
not matter, and do not have any impact on the macroeconomy.
Both Leiderman and Blejer (1988) and Seater (1993) provided in-depth overviews of the Ricardian
equivalence theorem. Surveys of previous empirical studies on Ricardian equivalence have been
produced by Gale and Orszag (2004), and Ricciuti (2003).
With little (recent) empirical knowledge on the efficacy of fiscal policy in modern economies, fiscal
stimulus policies have been enacted without a thorough understanding of the potency of these
policy actions – particularly given the marked structural changes in many developed economies
over the past two decades (such as the increased integration of global product and financial
markets). The analytical model employed in this paper considers the extent to which private saving
responds to changes in the total general government (Commonwealth, state and local) fiscal stance.
While this framework lends itself towards explaining Ricardian equivalence effects, it can also be
considered as a broad measure of the impact of fiscal policy on short- and long-run aggregate
demand. The model is estimated using the autoregressive distributed lag approach (ARDL) to
cointegration, which provides both short- and long-run coefficient estimates, but also provides the
flexibility to accommodate the introduction of coefficients for structural breaks.
However, it is likely that the Australian economy has been subject to a substantial amount of
structural change over the past 50 years. From the 1950s through to the early 1980s, the Australian
economy was heavily regulated, with markets subject to price controls and tariff protection, a fixed
exchange rate, and government controls on bank deposits, interest rates and credit. The 1980s saw a
period of rapid reform, with the floating of the dollar, removal of restrictions on credit creation,
interest rates, foreign capital inflows and other broader reforms around market pricing and removal
(or lowering) of tariffs and subsidies. Not accounting for these changes could lead to spurious
results in the econometric analysis. The Lee and Strazicich two-break unit root test is used to test
the time series properties of the data.
The following section discusses the analytical model to be estimated in this paper, along with the
expected signs of the explanatory variables. Section 3 uses unit root tests that allow for two
endogenously determined structural breaks in the individual time series. The analytical model is
then estimated through the ARDL approach to cointegration, which provides the flexibility to
incorporate structural breaks and both stationary and non-stationary time series. Conclusions are
presented in section 4.
2. Analytical framework
The relationship between private and public saving can be estimated through a model with the
following functional form:
5
t tZ e0 0 0priv pub
t tS Sα β φ= + + + (1)
where and denotes the ratio of net household plus net corporate saving (which gives total
net private saving) to GDP, and the ratio of net general (Commonwealth, local and state)
government saving to GDP, while
privtS pub
tS
tZ is a vector of control variables. This reduced-form saving
equation allows for the estimation of the private savings offset with a large number of control
variables, and is similar to that used in previous empirical studies by Haque (et al: 1999); Masson
(et al: 1998); Loayza (et al: 2000); Comley (et al: 2002); de Serres and Pelgrin (2003); and de Mello
(et al: 2004). A similar specification of this model was applied to the United States by Cotis (et al:
2006).
The vector tZ of control variables often includes conventional determinants of private saving, such
as the real interest rate, inflation, household income, social assistance payments to households,
changes in the terms of trade, and employment. Specifically:
{ }, , , , , , , ,t t t t t t t t t tZ Y AS U R INF TOT FLIB H EQ= (2)
Where:
= Household gross disposable income; tY
= Social assistance benefits to household gross disposable income; tAS
= Unemployment rate; tU
tR = Real interest rate;
tINF = Inflation rate;
= Terms of trade; tTOT
= Net foreign liabilities (proxy for financial openness); tFLIB
= Australian house price index (proxy for wealth); and tH
= Australian share price index (proxy for wealth). tEQ
The hypothesis of a strict private savings offset (Ricardian equivalence) would be supported if the
coefficient on public saving in (5.1) above, 0 1β = − , controlling for the other private saving
determinants. A negative coefficient on public savings, but less than 0, that is 0( 1 0)β− < < would
indicate a partial savings offset, and that changes in the general government sector’s fiscal stance
has measurable impacts on the wider economy.
Cotis (et al: 2006) discuss a number of reasons which could give rise to a positive coefficient on
public saving, that is, where 0 0β > . Sources of changes in the fiscal position arise not only from
changes to taxation arrangements, but also from changes in expenditures. For a positive private
savings offset, public expenditures need to be considered complimentary, with a clear distinction
between expenditures which are permanent, and those which are transitory. Permanent changes will
tend to generate negative private savings offsets through the restrictions imposed by the
intertemporal budget constraint. Temporary shocks in government spending, however, could
generate positive private saving responses, particularly when households see public and private
consumption as complements.1
1 Specifically, this arises when the marginal utility of private consumption is positively affected by public spending.
Government-subsidised health and education programmes, and government co-payment incentives for first home
buyers, could provide examples of public and private complements in consumption.
6
The coefficient on household disposable income, , is expected to be positive. As household
income may be considered a proxy for labour income in a standard life-cycle model of
consumption, an increase in household disposable income is expected to increase private saving.
Alternatively, households may suffer from consumption inertia and therefore take time to change
their consumption patterns to new levels of income.
tY
Social assistance payments to households, , are expected to negatively impact private savings.
The existence of a welfare safety net in Australia is expected to crowd out precautionary motives
for saving, and other privately-run alternatives that would encourage thrift.
tAS
Increasing levels of unemployment lowers disposable incomes, and, through a greater incidence of
liquidity constraints, lowers saving. However, increases in unemployment may increase the need for
precautionary saving. But as noted above the existence of welfare safety nets in Australia may
crowd out precautionary motives for saving. Overall the coefficient on the unemployment rate, ,
is expected to be of negative sign.
tU
The effects of inflation, tINF , and the real interest rate, tR , are somewhat ambiguous, and depend
largely on the extent of credit constraints and on the relative magnitude of income and substitution
effects. Also, higher, and/or accelerating inflation erodes the real value of debt and raises private
saving, but may also discourage holdings of assets that are not inflation-indexed.
Terms of trade shocks, , are particularly relevant for Australia given a high reliance on
commodity-based exports. This coefficient is expected to be positively correlated with private
saving to the extent that terms of trade shocks are viewed as being temporary
tTOT
2 through the Laursen-
Harberger-Metzler effect.3 Permanent shocks should not affect private saving.
As noted earlier, there has been a considerable amount of economic reform undertaken in Australia
over the past three decades, most notably the reform of Australia’s financial sector. Financial
liberalisation in Australia occurred over a decade beginning in the early 1980s, with removals of
restrictions on bank deposit rates and lending, and progressed to other significant reforms of which
the most notable were the floating of the Australian dollar in December 1983, and deregulation of 2 This historically has been the case with terms of trade shocks experienced with the Korean War, 1970s oil price
shocks, and most recently the rapid industrialisation of China. 3 According to the Laursen-Harberger-Metzler effect, an adverse (beneficial) transitory movement in the terms of trade
results in a decrease (increase) in a country’s current level of income which is larger than the decrease (increase) in its
permanent income, causing a fall (rise) in aggregate saving.
7
8
home mortgage interest rates. This period of financial deregulation lead to a marked structural shift
in the Australian economy and the development of sophisticated private markets for credit and
financial risk management. More sophisticated private credit markets also enables greater access to
personal credit, allowing households to smooth consumption.
As noted by de Mello (et al: 2004), the effect of financial liberalisation on private saving is
ambiguous, because improved access to credit may boost consumption but the removal of bank
portfolio allocation constraints, which often accompanies financial liberalisation, may result in
higher real interest rates, which encourages saving. Given the large increase in foreign capital
inflows following financial market deregulation, it may be reasonable to expect that any coefficient
representing financial openness in Australia will have a negative sign.
However, adequate proxies for financial openness are difficult to measure, and somewhat subjective
in nature. Proxies may include variables such as growth in M2 money and the ratio of household
wealth to disposable income (as used by Comley et al: 2002). However, long time series for these
variables are generally not available, with most measures only dating back to around the early
1980s at best. Alternative measures of financial openness have been suggested by Lane and Milesi-
Ferretti (2001), and include measures based around countries’ foreign assets and liabilities. Given
this, Australia’s level of net foreign liabilities may provide a good proxy for financial openness,
particularly as foreign debt has increased substantially since the financial market reforms of the
1980s. Data on Australia’s net foreign liabilities is also available back to the late 1950s.
Household wealth is expected to affect consumption/saving decisions based on permanent income
considerations. Given that most Australian households have historically tended to hold their wealth
through the family home, a house price index is used here as it is expected to provide a good proxy
for household wealth in Australia.4
A share price index is also considered as an additional measure of private wealth. Historically, the
proportion of Australian households participating directly in the sharemarket had been relatively
low – until rising markedly over the past two decades. In 2006, approximately 38 per cent of the
Australian population owned shares directly (Australian Securities Exchange: 2007),5 which places
Australia as having some of the highest (direct) share ownership rates in the world.
4 Around 70 per cent of Australian households owned their home in 2003-04 (Australian Bureau of Statistics: 2006). 5 Australian households have also been undertaking greater ownership of equities indirectly through their
superannuation savings. The Australian Securities Exchange (2007) estimates that in 2006, approximately 46 per cent of
Data has been sourced from the Australian Bureau of Statistics and the Reserve Bank of Australia.
The sample size is large in both the number of observations (192) and the time period which is
considered: 1959:3 – 2007:2.
3. Econometric methodology and empirical testing
Unit root tests
Previous studies which examine both Ricardian equivalence and fiscal multipliers usually have
examined the time series properties of variables by using the Augmented Dickey-Fuller (ADF)
(1979, 1981) or Philip-Perron (1988) unit root tests. However, these tests do not allow for the
possibility of one or more structural breaks in the time series. Perron (1989) argued that in the
presence of a structural break, the standard ADF tests are biased toward the non-rejection of the null
hypothesis. The timing of the structural break in Perron’s procedure is assumed to be known a
priori in accordance with underlying asymptotic distribution theory. Test statistics are constructed
by adding dummy variables representing different intercepts and slopes, thereby extending the
standard ADF procedure.
However, Perron’s technique was criticised by Christiano (1992) as specific break-dates may be
chosen which support the researcher’s results and a priori expectations (i.e. data mining). Since
then, a number of studies have been developed using different methodologies for endogenising the
structural breaks. These studies include Banerjee (et al: 1992), Zivot and Andrews (1992), Perron
(1997) Lumsdaine and Papell (1998), and Lee and Strazicich (2003).
Lee and Strazicich (2003) developed a two-break minimum Lagrange Multiplier (LM) unit root test
where the alternative hypothesis implies trend stationarity (referred to by the authors as “trend-
break stationarity”).6 This test allows for up to endogenous structural breaks, which may occur in
either the level or slope of a series. First consider the following data-generating process:
9
t 't ty Z eδ= + (3)
1t te e tuβ −= + (4)
the Australian population owned shares either directly via shares or indirectly via a managed fund or self managed
superannuation fund. 6 The null hypothesis is a unit root with breaks.
where ty is the data series in period , t δ is a vector of coefficients, tZ is a matrix of exogenous
variables, and is a standard white noise error term with zero mean and constant variance tu
2(iid ,t ) 0u N σ∼ , tZ is described by , to allow for a constant term, linear
time trend, and two structural breaks in level and trend where denotes the time period of the
breaks. Under the trend-break stationary alternative, the
'* *1 2 1 2, , ,t t t tt D D DT DT⎡⎣1, , ⎤⎦
BjT
jtD
1Bj≥ + 1, 2j
terms describe an intercept shift in the
deterministic trend, where for , 1jtD = t T = , and zero otherwise; jtDT describes a
change in slope of the deterministic trend, where 1jtDT = for , , and zero
otherwise.
1+Bjt T≥ 1, 2=j
The two-break minimum LM unit root test statistic is obtained from the following regression:
'1t t t i t iy d Z S y S tφ ε− −Δ = Δ + + Δ +∑ (5)
where t t x tS y Zψ δ= − − 2,...,t T=, and 1 1t y Zψ δ= − . is a de-trended series of tS ty using the
coefficients in tδ , which are estimated from the regression in first differences of tyΔ on
[ ]1 21, , ,t t t 1 2t t,Z D D DTΔ Δ Δ DTΔ 1Δ = , y and 1Z are the first observations of ty and tZ , respectively,
and Δ is the first difference operator. The standard white noise error terms is represented by tε . To
correct for serial correlation, , 1tS −Δ 1,...,I k= terms are included. The unit root hypothesis in
equation (5) is equivalent to 0φ = , and the test statistics are defined as:
Tρ φ= ⋅ (6)
τ = t-statistic for the null hypothesis 0φ = . (7)
To determine (endogenously) the location of the two breaks ( )/ , 1, 2j BjT T jλ = = , the minimum
LM unit root test uses a grid search procedure:
( )LM Infρ λ ρ λ= (8)
( )LM Infτ λτ λ= (9)
10
The LM test is corrected for autocorrelated errors by including lagged augmentation terms
as per the standard Augmented Dickey-Fuller test. The optimal lag length, k , is
determined through the general to specific procedure of Perron (1989).
, 1,...,St j j kΔ − =
The Lee and Strazicich two-break LM unit root test was conducted in GAUSS using code provided
by the authors. Again Models A and C were run, with lag lengths generated automatically through a
general to specific procedure.
11
)
Critical values for the two-break LM unit root test also vary depending on the location of the breaks
1 2( / , /B BT T T Tλ = and are symmetric around λ and (1 )λ− . Critical values for the two-break
minimum LM unit root test7 for Model C (intercept and trend break) are shown in Table 1 below,
and are drawn from Table 2 in Lee and Strazicich (2003). Critical values for the two-break LM unit
root test with change in intercept (Model A) at the 1%, 5% and 10% levels respectively are -4.55,
-3.84, and -3.50.
Table 1: Critical values for the two-break LM unit root test (Model C)
Break points λ = (TB1/T, TB2/T) Critical values
1% 5% 10%
λ = (0.2,0.4) -6.16 -5.59 -5.27
λ = (0.2,0.6) -6.41 -5.74 -5.32
λ = (0.2,0.8) -6.33 -5.71 -5.33
λ = (0.4,0.6) -6.45 -5.67 -5.31
λ = (0.4,0.8) -6.42 -5.65 -5.32
λ = (0.6,0.8) -6.32 -5.73 -5.32
Results from Model A (Table 2) suggest that with the exception of private saving (PS), all of the
variables contain a unit root with at least on statistically significant structural break.
7 Critical values are provided by Lee and Strazicich for T = 100. Unfortunately the authors do not provide critical values
for larger or smaller sample sizes.
12
Table 2: Results of the two-break LM unit root test (Model A) Variable k TB T φ = 0 Inference
lnPS 0 1997:4#, 2001:1# -5.4227* Stationary
GS 4 1976:2#,1999:2# -3.4204 Non-Stationary
lnY 8 1966:2,1987:3# -1.7050 Non-Stationary
lnFLIB 7 1971:4#,1976:4# -3.3650 Non-Stationary
lnU 4 1971:4,1974:4 -2.1289 Non-Stationary
R 4 1977:3#,1983:4# -3.0836 Non-Stationary
INF 8 1975:3#,1983:2 -2.2589 Non-Stationary
lnAS 7 1992:1#, 1998:3# -2.8172 Non-Stationary
lnTOT 7 1974:1#,1974:3 -2.4932 Non-Stationary
lnH 2 1973:3, 1980:4# -1.9984 Non-Stationary
LnEQ 3 1983:2, 1988:1# -3.3574 Non-Stationary A maximum of 8 lags was specified in GAUSS. # Denotes significance at the 5% level for the break-point dummy variables. Critical value for T φ = 0 is -3.84 at the 5% level. * Denotes significance at the 5% level.
When allowing for a break in both the level and trend of the series, Model C (Table 3) produces
quite different results. In contrast to Model A, the results in Table 3 suggest that household
disposable incomes (Y), inflation (INF), and the terms of trade (TOT) are also stationary series.
Table 3: Results of the two-break LM unit root test (Model C) Variable k TB T φ = 0 Critical value
break points Inference
lnPS 0 1997:4#, 2001:1 -6.5213* λ = (0.8,0.9) Stationary
GS 7 1974:3#,1997:2# -4.8116 λ = (0.3,0.8) Non-Stationary
lnY 6 1973:2#, 1992:3 -6.7481* λ = (0.3,0.7) Stationary
lnFLIB 8 1973:1#,1986:1# -4.3292 λ = (0.2,0.7) Non-Stationary
lnU 6 1974:2#,1988:1# -4.5601 λ = (0.3,0.6) Non-Stationary
R 4 1973:2,1985:3 -4.9872 λ = (0.3,0.6) Non-Stationary
INF 7 1973:2#,1991:4# -6.6046* λ = (0.3,0.7) Stationary
lnAS 7 1970:1, 1976:1# -5.4113 λ = (0.2,0.4) Non-Stationary
lnTOT 4 1969:4#,1995:4# -6.0485* λ = (0.2,0.8) Stationary
lnH 2 1972:2#, 1993:1# -3.9289 λ = (0.3,0.7) Non-Stationary
lnEQ 3 1973:2#, 1986:4# -5.2620 λ = (0.3,0.6) Non-Stationary
A maximum of 8 lags was specified in GAUSS. # Denotes significance at the 5% level for the break-point dummy variables. Critical values for T φ = 0 are contained in Table 1. * Denotes significance at the 5% level.
When interpreting results from the LM unit root tests, the timing of structural breaks could be a
useful guide for discerning the reliability and effectiveness of the procedure. Judgement of each
model (A or C) based upon economic theory and historical events, such as policy changes and
economic shocks (for example), can help to determine the timing of structural breaks, and whether
these changes have been sudden or gradual. Results indicate that structural changes have generally
coincided with a number of significant events over the past few decades, including:
• the 1960s resources boom;
• the expansion of social welfare programmes (Whitlam Government);
• oil price (terms of trade) and inflation shocks in the 1970s;
• the extensive period of financial deregulation in the 1980s; and
• the 1990-91 recession.
Cointegration
Conventional cointegration procedures (such as that of Johansen (1991, 1995), usually require that
all data entering into an equation be non-stationary. As the unit root tests undertaken above suggest
that the ratio of private saving to GDP is a stationary time series, conventional cointegration
techniques cannot be used to estimate the analytical model. Further, the unit root tests also
suggested that each data series contains at least one structural break. This further complicates the
use of cointegration techniques as conventional cointegration methods cannot account for
endogenous structural breaks. While recent econometric developments allow for cointegration
testing in the presence of structural breaks, these techniques are currently in their early stages of
development and often can only accommodate one structural break (earlier techniques such as that
of Gregory and Hansen (1996) also require all data to be non-stationary). To overcome these
difficulties, the analytical model will be estimated through the autoregressive distributed lag
(ARDL) approach to cointegration (see Pesaran and Shin 1998; Pesaran et al 1996; and Pesaran et
al 2001). This technique allows for a greater degree of flexibility – allowing for both stationary and
non-stationary data – and can accommodate additional variables that can represent structural breaks.
Following Pesaran (et al: 2001) the ARDL technique involves two steps for estimating the
cointegrating relationship. Under the first step, the existence of a long-run cointegrating relationship
is tested. If a long-run cointegrating relationship is found, the second step involves estimating both
the long and short-run coefficients. An intercept and trend term will be added to the estimation of
the model – particularly as the united root tests considered in the previous section indicated that the
dependent variable (PS) is stationary – and a visual inspection of the ratio of private saving to GDP
indicates a considerable downward trend in the data series. Therefore, the ARDL model is a general
ECM with unrestricted intercept and trend:
13
t t 1
'0 1 1 . 1
1'
p
t yy t yx x t i t ii
y a a t y x z w xπ π−
− − −=
Δ = + + + + Ψ Δ + Δ +∑ ε (10)
where and . As noted above, the first step of the ARDL procedure involves testing for
a cointegrating relationship. This step tests for the absence of any level relation between
0 0a ≠ 1 0a ≠
ty and tx
via the exclusion of the lagged level variables 1ty − and 1tx − in equation (7.6). Persaran (et al: 2001)
define the F-statistic tests for the null hypotheses as 0 : 0yyyy
π πH = , ' and the
alternative hypotheses as ,
.0 .:yx x
yx xH π π = 0
1 : 0yyyyH π π ≠ .
1yx xπ
.: yx xH π 0 '≠ . The joint null hypothesis for (10) is
given by:
(11) .0 0 0
yy yx xH H Hπ π= ∩
and the alternative hypothesis is correspondingly stated as:
(12) .1 1 1
yy yx xH H Hπ π= ∪
The asymptotic distribution of the F-statistics are non-standard under the null hypothesis of no
cointegrating relationship between the variables, regardless of the order of integration of the
variables being considered. The calculated F-statistic is compared with the critical values provided
in Pesaran (et al: 2001). The null hypothesis of no cointegration is rejected if the calculated F-
statistic is greater than the upper bound critical value. If the calculated F-statistic falls below the
lower bound, then the null hypothesis of no cointegration cannot be rejected. The result is
inconclusive if the calculated F-statistic lies between the upper and lower bound critical values.
The ARDL specification for equation (1) is as follows:
0 11 1 1 1
p p p p
t i t i i t i i t i ii i i i
PS t PS GS Y ASα α δ β φ ϕ− − −= = = =
Δ = + + Δ + Δ + Δ + Δ +∑ ∑ ∑ ∑ t i−
− +
− +
t
1 1 1 1 1
p p p p p
i t i i t i i t i i t i i t ii i i i i
U R INF TOT FLIBγ τ υ ρ ψ− − − −= = = = =
Δ + Δ + Δ + Δ + Δ∑ ∑ ∑ ∑ ∑
1 1 2 1 3 1 4 1 5 11 1
p p
i t i i t i t t t t ti i
H EQ PS GS Y AS Uξ ω λ λ λ λ λ− − − − − −= =
Δ + Δ + + + + +∑ ∑
6 1 7 1 8 1 9 1 10 1 11 1t t t t t tR INF TOT FLIB H EQ uλ λ λ λ λ λ− − − − − −+ + + + + + (13)
where and have been shortened to PS and GS respectively. In the ARDL specification
above, the summation signs represent the error correction dynamics, while the second section of the
equation, denoted by
privtS pub
tS
iλ , represents the long run relationship. The null hypothesis of no
cointegration in equation (13) is given by:
14
0 1 2 3 4 5 6 7 8 9 10 11: 0H λ λ λ λ λ λ λ λ λ λ λ= = = = = = = = = = =
or equivalently as:
( , , , , , , , , ,PSF PS GS Y AS U R INF TOT FLIB H EQ)
The corresponding alternative hypothesis is:
1 2 3 4 5 6 7 8 9 10 110, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0λ λ λ λ λ λ λ λ λ λ λ≠ ≠ ≠ ≠ ≠ ≠ ≠ ≠ ≠ ≠ ≠
As noted earlier, the relevant test statistic here is the F-statistic for the joint significance of the
coefficients, and as we are dealing with quarterly data, a maximum of 4 lags is included.
Table 4: Results from bounds test on equation (13) – 1959:3 to 2006:2 Dep. Var. F-statistic Probability Conclusion
( , , , , , , , , ,PS )F PS GS Y AS U INF R TOT FLIB H EQ 3.4906* 0.000 Cointegration
( , , , , , , , , ,GS )F GS PS Y AS U INF R TOT FLIB H EQ 2.4126 0.009 Inconclusive
( , , , , , , , , ,Y )F Y PS GS AS U INF R TOT FLIB H EQ 2.2677 0.015 No cointegration
( , , , , , , , , ,AS )F AS PS GS Y U INF R TOT FLIB H EQ 2. 4465 0.008 Inconclusive
( , , , , , , , , ,U )F U PS GS Y AS INF R TOT FLIB H EQ 3.0196 0.001 Inconclusive
( , , , , , , , , ,R )F R PS GS Y AS U INF TOT FLIB H EQ 2.1676 0.020 No cointegration
( , , , , , , , , ,INF )F INF PS GS Y AS U R TOT FLIB H EQ 2. 0838 0.026 No cointegration
( , , , , , , , , ,TOT )F TOT PS GS Y AS U INF R FLIB H EQ 3.5018* 0.000 Cointegration
( , , , , , , , , ,FLIB )F FLIB PS GS Y AS U INF R TOT H EQ 1.7875 0.063 No cointegration
( , , , , , , , , ,H )F H PS GS Y AS U INF R TOT FLIB EQ 3.1870 0.001 Inconclusive
( , , , , , , , , ,EQ )F EQ PS GS Y AS U INF R TOT FLIB H 1.8996 0.045 No cointegration
Asymptotic critical value bounds are obtained from Table CI(iii), Case V: unrestricted intercept and unrestricted trends for k=10 (Persaran et al: 2001). Lower bound I(0)=2.33 and Upper bound I(1)=3.46 at the 5% significance level. * Denotes significance at the 5% level. ** Denotes significance at the 1% level. Where private savings is the dependent variable, the calculated F-statistic of 3.4906 is greater than
the upper bound critical value at the 5 per cent level, which rejects the null hypothesis of no
cointegration – implying a long-run level relationship between the variables (Table 4). Considering
the possibility of reverse causation, where government savings is the long-run dependent variable,
the calculated F-statistic of 2.4126 falls into the inconclusive region. Consequently, reverse
causation cannot be ruled-out. Where the cointegration tests are undertaken with different
dependent variables, the results also suggest a long-run relationship between the variables, and that
15
16
Y, R, INF, FLIB, and EQ act as the long-run forcing variables for private saving. While results in
Table 1 show inconclusive results for social assistance payments (AS), unemployment (U), and
house prices (H), the subsequent estimations of the short- and long-run parameters may yield
further information on the significance of these variables.
The structural breaks identified above may be accounted for by the inclusion of break-point dummy
variables in the ARDL model. The structural breaks to be included in the ARDL specification are:
B1969:1 = 1960s resources boom;
B1973:3 = expansion of social welfare programmes (Whitlam Government); oil price shocks
and inflation8;
B1984:1 = floating of the Australian dollar9, including broader financial market liberalisation;
and
B1990:1 = onset of recession in the early 1990s.
Estimation results
The estimated long-run coefficient estimates for equation (13) are provided in Table 5.10 With the
exception of the unemployment rate (U), all variables have the expected sign, although the wealth
variables will be discussed in greater detail below. For the level of government savings (GS), the
results suggest that over the long run, changes in general government saving are offset by changes
in private savings by almost half (-0.44). This implies that the behavioural response of households
and corporations is not fully Ricardian, and that fiscal policy has a (partial) flow through to the real
economy – potentially impacting output, real interest rates, the exchange rate, and subsequently the
current account. The value of this coefficient is also similar to the results of Comley (et al: 2002),
who estimated a long-run private savings offset coefficient for Australia of -0.5. However, it is
important to note here that Comley’s estimated long-run coefficient was not statistically significant,
possibly due to having a much smaller sample (1981:1-2002:2).
8 While two breaks may have been included for each of these effects, the close proximity of both breaks would mean
that the inclusion of separate dummy variables for each could increase the likelihood of serial correlation in the
regression estimates. 9 The floating of the Australian dollar is considered to be the most significant of the broader financial market reforms
undertaken over the decade from the late 1970s though to the late 1980s. 10 The appropriate lag length was chosen according to the Schwarz Bayesian Criterion.
17
The estimated Australian private savings offset of -0.44 is however lower than some estimates
derived through international panel studies. Considering private savings across a panel of 21 OECD
countries, de Mello (et al: 2004) estimated a long-run private savings offset coefficient of around
-0.75; implying that changes in the fiscal stance are almost fully offset by corresponding changes in
private saving. Following an analytical model similar to that used here, and to that employed by de
Mello (et al: 2004), Cotis (et al: 2006) estimated a long-run private savings offset of around two
thirds for a panel of 16 OECD countries. Isolating impacts on the United States, Cotis (et al: 2006)
estimated a positive long-run private savings coefficient – implying that US households behave in a
non-Ricardian manner.11
Table 5: Estimated long-run coefficients for equation (13) ARDL (1,0,1,2,0,0,0,0,0,0,0) selected lags based on Schwarz Bayesian Criterion
Variable Coefficient Standard Error T-Ratio Probability
Constant -0.2564 0.1157 -2.2152* 0.028
Trend 0.0003 0.0003 0.9729 0.332
GS -0.4438 0.1178 -3.7673** 0.000
Y 0.4241 0.1409 3.0100* 0.003
U 0.1571 0.2082 0.7542 0.452
R 0.0301 0.0729 0. 4128 0.680
INF -0.1460 0.1094 -1.3340 0.184
AS -0.4579 0.2145 -2.1342* 0.034
TOT 0.0008 0.0002 3.9830** 0.000
FLIB -0.0364 0.0155 -2.3410* 0.020
H -0.0066 0.0127 -0.5153 0.607
EQ 0.0179 0.0106 1.6806 0.095
B1969 0.0029 0.0062 0.4685 0.640
B1973 -0.0161 0.0106 -1.5082 0.133
B1984 -0.0035 0.0066 -0.5388 0.591
B1990 -0.0151 0.0078 -1.9209 0.056 * Denotes significance at the 5% level. ** Denotes significance at the 1% level.
For the remaining variables in Table 5, the results indicate that for a 1 per cent rise in household
gross disposable income (Y), the ratio of private savings to GDP increases by 0.42 per cent. This
also implies a marginal propensity to consume of approximately 0.6 – which is consistent with
Australian National Accounts data (which indicates that the consumption share of GDP in Australia 11 Changes in public saving result from both taxation and expenditure. While permanent expenditures will generate an
increase in private saving through the intertemporal budget constraint, temporary expenditure shocks can generate
positive private saving offsets (particularly when households see public and private consumption as complements; for
example, rebates and co-payments).
of 60 per cent). Rising levels of social assistance payments to households (AS) are estimated to have
a negative impact on private savings over the long-run, with the ratio of private saving to GDP
declining by around 0.46 per cent for each one per cent increase in social assistance payments to
households. Australia’s terms of trade are (TOT) is estimated to have a small, although statistically
significant, positive impact on private savings over the long run. As expected, financial
liberalisation has a negative impact on private savings over the long run. For the unemployment rate
(U), the real interest rate (R), and inflation (INF), the results in Table 5 indicate that these variables
do not have a statistically significant long-run impact on the level of private saving in Australia.
Both of the wealth variables present some interesting results. Changes in the prices of household
assets (and the returns derived from these) will affect household consumption and saving.
Additionally, as the dependent variable is private saving (which includes corporate saving), changes
in wealth will also affect business borrowing and investment decisions. Results here indicate that
wealth from housing does not exert a statistically significant impact on private saving over the long
run, although it is of the expected sign. Given that most Australian’s hold wealth through the family
home, this is somewhat surprising. Equity prices appear to have had a statistically significant (albeit
at the 10 per cent level) impact on private saving over the long run. The positive sign of this
coefficient is curious, and suggests that for a 1 per cent rise in equity prices, the ratio of private
saving to GDP rises by around 0.02 per cent. This positive response may be somewhat indicative of
the broad shift toward equity investment, particularly the indirect investment occurring through
households’ accumulation of assets in superannuation.
Of the dummy variables included in the estimation, only the structural break coinciding with the
early 1990s recession (B1990) is estimated to have had a statistically significant (at the 10 per cent
level) long-run impact on the private savings ratio.
The short-run error correction estimates are presented in Table 6. In the short-run, the error
correction equation indicates a private saving offset of one quarter (-0.25) to changes in government
saving. The error correction term, ( 1)ecm − , is of the correct sign and statistically significant –
indicating that deviations from the long-run rate of private saving are corrected by over 50 per cent
in the next period, which is a relatively fast pace of adjustment back to equilibrium. While the
unemployment rate (U) was statistically insignificant in the long-run relationship, the estimated
coefficient here is of the correct sign, and significant at the 10 per cent level, whilst the lagged value
of unemployment is significant at the 1 per cent level. This suggests that the unemployment rate
negatively impacts private saving in the short-run only, which would be consistent with the impact
of temporary shocks to output.
18
Table: 6 Error correction representation of equation (13) ARDL (1,0,1,2,0,0,0,0,0,0,0) selected lags based on Schwarz Bayesian Criterion
Variable Coefficient Standard Error T-Ratio Probability
Constant -0.1469 0.0692 -2.1224* 0.035
Trend 0.0002 0.0002 0.9838 0.327
GSΔ -0.2544 0.0675 -3.7637** 0.000
YΔ 0.5249 0.0747 7.0231** 0.000
UΔ -0.3919 0.2228 -1.7593 0.080
( 1)UΔ − -0.7711 0.2184 -3.5302** 0.001
RΔ 0.0172 0.0419 0.4119 0.681
INFΔ -0.0804 0.0593 -1.3568 0.177
ASΔ -0.2624 0.1208 -2.1718* 0.031
TOTΔ 0.0004 0.0001 3.8787** 0.000
FLIBΔ -0.0208 0.0086 -2.4049* 0.017
HΔ -0.0037 0.0072 -0.5176 0.605
EQΔ 0.0102 0.0060 1.7059 0.090
1969BΔ 0.0016 0.0036 0.4645 0.643
1973BΔ -0.0092 0.0061 -1.5230 0.130
1984BΔ -0.0020 0.0038 -0. 5415 0.589
1990BΔ -0.0087 0.0047 -1.8481 0.066
( 1)ecm − -0.5732 0.0597 -9.6020** 0.000
* Denotes significance at the 5% level. ** Denotes significance at the 1% level. 0.444 * 0.424 * 0.157 * 0.03 * 0.160 * 0.458 * 0.0007 * 0.036 *ecm PS GS Y U R INF AS TOT FLIB= + − − − + + − + +
0.007 * 0.018 * 0.256 * Constant 0.0003 * Trend 0.003 * 1969 0.016 * 1973 0.004 * 1984 0.015 * 1990H EQ B B B B− + − − + + +
20.6249R =
20.5844R = F-stat [ ](17,168) 17.3865 0.000F = SER 0.0082=
RSS 0.011= DW-statistic 2.0817=
Short-run coefficient estimates for household gross disposable income (Y), and the terms of trade
(TOT) are significant at the 1 per cent level, while social assistance payments (AS), and financial
openness (FLIB) are significant at the 5 per cent level. Similar to the long-run results, the estimated
short-run coefficients for the real interest rate (R), inflation (INF) and break-point dummy variables
B1969, B1973, and B1984 are statistically insignificant. The short-run results also indicate that
housing wealth is not statistically insignificant, while wealth from equities appears to bear a
statistically significant influence (at the 10 per cent level) on the ratio of private saving to GDP in
Australia.
Diagnostic statistics from the estimations are positive (Table 7), indicating that the error terms do
not suffer from serial correlation, and are normally distributed. The model specification also
satisfies the RESET test for omitted variables and functional form.
19
Table 7: Diagnostic tests on equation (13) LM Test Statistics 2χ statistic Probability
Serial correlation a 2(4)χ 3.3784 0.497
Normality b 2(2)χ 1.5196 0.468
Functional form c 2(1)χ 0.0038 0.951
Heteroscedasticity d 2(1)χ 0.0179 0.893
* Denotes significance at the 5% level. ** Denotes significance at the 1% level. a Breusch-Godfrey LM test for serial correlation. b Jarque-Bera normality test. c Ramsey RESET test for omitted variables/functional form. d White test for heteroscedasticity.
Two subsample estimations for equation (13) will now be undertaken. These cover the period
1959:3 – 1983:4, while the second period is over 1984:1 – 2006:2. This will attempt to account for
the effects of financial market liberalisation, and a move toward a greater integration of the
Australian economy into the global financial system – particularly as the break-point dummy
variable (B1984) was not statistically significant in the earlier analysis.12
Over the first subsample period, the Australian economy was highly regulated, with a fixed
exchange rate, tariff controls, and other regulations over the financial system such as controls on
bank lending, deposits, and some interest rates (such as mortgage interest rates, overnight money
market rates, and deposit rates). Since the floating of the Australian dollar and associated financial
market reforms, foreign capital inflows into Australia have increased markedly, and there has been
a commensurate increase in financial market innovation. This integration into global capital markets
may have dampened the impact of fiscal policy on the economy. These reforms have also occurred
in concert with other reforms in the labour market, tariff reform, the establishment of free trade
arrangements with some countries, a national competition policy agenda, fiscal consolidation,
privatisation of government business enterprises, and the introduction of inflation targeting.
Private saving offsets – 1959:3 to 1983:4
Cointegration tests where private saving (PS) is the dependent variable yield an F-statistic of
3.7095, which is greater than the upper bound critical value at the 5 per cent level – implying that
the long-run level relationship between these variables is still observed over the first subsample
period (Table 8). However, where government savings is the dependent variable, the calculated F-
statistic again falls into the inconclusive zone.
12 As the financial reforms were phased over the 1980s, with the floating of the Australian dollar one of several major
reforms, the insignificance of this dummy variable is not that surprising. This implies that a gradual structural change
may have been occurring as opposed to a sudden level shift.
20
Table 8: Results from bounds test on equation (13) – 1959:3 to 1983:4 Dep. Var. F-statistic Probability Conclusion
( , , , , , , , , ,PS )F PS GS Y AS U INF R TOT FLIB H EQ 3.7095* 0.001 Cointegration
( , , , , , , , , ,GS )F GS PS Y AS U INF R TOT FLIB H EQ 2.5843 0.016 Inconclusive
( , , , , , , , , ,Y )F Y PS GS AS U INF R TOT FLIB H EQ 1.1575 0.349 No cointegration
( , , , , , , , , ,AS )F AS PS GS Y U INF R TOT FLIB H EQ 3.2765 0.003 No cointegration
( , , , , , , , , ,U )F U PS GS Y AS INF R TOT FLIB H EQ 2.1103 0.045 No cointegration
( , , , , , , , , ,R )F R PS GS Y AS U INF TOT FLIB H EQ 2.1373 0.043 No cointegration
( , , , , , , , , ,INF )F INF PS GS Y AS U R TOT FLIB H EQ 1.6689 0.121 No cointegration
( , , , , , , , , ,TOT )F TOT PS GS Y AS U INF R FLIB H EQ 2.4355 0.022 Inconclusive
( , , , , , , , , ,FLIB )F FLIB PS GS Y AS U INF R TOT H EQ 2.2704 0.032 No cointegration
( , , , , , , , , ,H )F H PS GS Y AS U INF R TOT FLIB EQ 2.7366 0.011 Inconclusive
( , , , , , , , , ,EQ )F EQ PS GS Y AS U INF R TOT FLIB H 3.7878 0.001 Cointegration
Asymptotic critical value bounds are obtained from Table CI(iii), Case V: unrestricted intercept and unrestricted trends for k=10 (Persaran et al: 2001). Lower bound I(0)=2.43 and Upper bound I(1)=3.56 at the 5% significance level. * Denotes significance at the 5% level. ** Denotes significance at the 1% level.
For the ARDL estimation over the period 1959:3-1983:4, initial results for equation (13) were not
positive, and indicated that the errors of the estimated ARDL were serially correlated and not
normally distributed. Additionally, the estimated trend coefficient was of the wrong sign. The trend
coefficient was dropped, along with estimated coefficients for the real interest rate (R), inflation
(INF), financial openness (FLIB), and the break-point dummy variables (B1969) and (B1973) as
these variables were all statistically insignificant. Serial correlation was still apparent in the model,
and despite theory suggesting that wealth effects may explain some of the variation in private
saving behaviour; both the house and equity price series were also dropped from the model.
Removing these improved the results markedly, with the Jarque-Bera test indicating that the
residuals were normally distributed, while the Breusch-Godfrey LM test suggested that serial
correlation had also been alleviated. This left the following specification for the subsample ARDL:
01 1 1 1
p p p p
t i t i i t i i t i ii i i i
PS PS GS Y Uα δ β φ γ− − −= = = =
Δ = + Δ + Δ + Δ + Δ∑ ∑ ∑ ∑ t i− +
− +
tu
1 1 2 11 1
p p
i t i i t i t ti i
AS TOT PS GSϕ ρ λ λ− − −= =
Δ + Δ + +∑ ∑
3 1 4 1 5 1 6 1t t t tY U AS TOTλ λ λ λ− − − −+ + + + (14)
21
The estimated long-run coefficient estimates for equation (14) are provided in Table 9. For the ratio
of government saving to GDP (GS) over the period 1959:3-1983:4, the estimated coefficient is
-0.39, which is somewhat lower than the full sample estimation. This potentially suggests that with
a lower private saving offset, fiscal policy may have exerted a larger impact on the real economy
during this period. Such a result would be consistent with the structure of the economy at that time
(markets being subject to a greater degree of regulation, and less exposure to international capital
and price movements) and confirms a priori expectations regarding these policy impacts.
Table 9: Estimated long-run coefficients for equation (14) ARDL (1,0,1,0,2,0) selected lags based on Schwarz Bayesian Criterion
Variable Coefficient Standard Error T-Ratio Probability
Constant -0.2085 0.0648 -3.2159** 0.002
GS -0.3994 0.1861 -2.1455* 0.035
Y 0.3906 0.0700 5.5746** 0.000
U -0.1998 0.2475 -0.8075 0.422
AS -0.2438 0.2855 -0.8539 0.395
TOT 0.0007 0.0003 2.6296** 0.010 * Denotes significance at the 5% level. ** Denotes significance at the 1% level.
A one per cent rise in household gross disposable income (Y) is estimated to raise the ratio of
private saving to GDP by 0.39 per cent over the first subsample, which is slightly higher than for
the full sample estimation. The terms of trade (TOT) is statistically significant, but is estimated to
only exert an extremely small impact on the private saving to GDP ratio. As expected, over this
subsample the ratio of social assistance payments to household gross disposable income (AS) and
the unemployment rate (U) are estimated to have had a statistically insignificant long-run impact on
private saving.
The short-run error correction estimates are presented in Table 10. In the short-run, the error
correction equation indicates a private saving offset of -0.23. The error correction term, ( 1)ecm − , is
of the correct sign and statistically significant – indicating that deviations from the long-run rate of
private savings are corrected by over 50 per cent in the next period. Household gross disposable
income, (Y), is statistically significant (at the one per cent level) while the estimated coefficient for
social assistance payments (AS) is markedly higher in the short-run, and includes an additional lag
coefficient for adjustment. The larger sign of this coefficient in the short run may again be
explained by the steep rise in the unemployment rate in 1974, then rising again in 1983 (where the
unemployment rate reached 10.2 per cent in the September quarter 1983) – suggesting that
households were more dependent on the welfare safety net over this period. However, it is
interesting that the results indicate that the unemployment rate is statistically insignificant in both 22
the long and short-run estimations. Prior to the large rise in unemployment during the 1970s, the
unemployment rate averaged 2 per cent over the 1960s. The introduction of expanded social welfare
programmes by the Whitlam government almost coincided with a steep rise in unemployment in
1974, which may explain this curio.13
Table 10: Error correction representation of equation (14) ARDL (1,0,1,0,2,0) selected lags based on Schwarz Bayesian Criterion
Variable Coefficient Standard Error T-Ratio Probability
Constant -0.1216 0.0413 -2.9407* 0.004
GSΔ -0.2329 0.1021 -2.2812* 0.025
YΔ 0.4916 0.0806 6.0980** 0.000
UΔ -0.1165 0.1462 -0.7968 0.428
ASΔ -1.4175 0.3407 -4.1602** 0.000
( 1)ASΔ − -0.7800 0.3133 -2.4892* 0.015
TOTΔ 0.0004 0.0002 2.6569* 0.009
( 1)ecm − -0.5831 0.0945 -6.1691** 0.000
* Denotes significance at the 5% level. ** Denotes significance at the 1% level. 0.399 * 0.391* 0.199 * 0.244 * 0.0007 * 0.209 * Constantecm PS GS Y U AS TOT= + − + + − +
20.7104R =
20.6800R = F-stat [ ](7, 88) 30.1357 0.000F = SER 0.0078=
RSS 0.0053= DW-statistic 1.9847=
Diagnostic statistics for the error correction mechanism (Table 11) are positive and indicate that the
model is correctly specified. The error terms are normally distributed and the Breusch-Godfrey LM
test indicates that no serial correlation is present.
Table 11: Diagnostic tests on equation (14) LM Test Statistics 2χ statistic Probability
Serial correlation a 2(4)χ 2.8417 0.585
Normality b 2(2)χ 3.7570 0.153
Functional form c 2(1)χ 0.6502 0.420
Heteroscedasticity d 2(1)χ 0.4577 0.499
* Denotes significance at the 5% level. ** Denotes significance at the 1% level. a Breusch-Godfrey LM test for serial correlation. b Jarque-Bera normality test. c Ramsey RESET test for omitted variables/functional form. d White test for heteroscedasticity.
13 In the absence of social welfare arrangements, the coefficient on unemployment could in fact be positive; inferring
that a rise in unemployment spurs an increase in precautionary saving.
23
Private saving offsets – 1984:1 to 2006:2
Cointegration tests where private saving is the dependent variable yield an F-statistic of 2.766,
which falls within the inconclusive range of the critical values at the 5 per cent level (Table 12).
Results from the bounds test also suggest reverse causation where government savings is the
dependent variable. Given the overall sample results presented earlier lend support to cointegration
the ARDL estimations will still be undertaken. The inconclusive result (and the suggested reverse
causation with government savings as the dependent variable) may in fact suggest that financial
liberalisation in Australia, leading to deeper and more open capital markets, has eroded the
transmission of changes in the government’s fiscal stance.
Table 12: Results from bounds test on equation (13) – 1984:1 to 2006:2 Dep. Var. F-statistic Probability Conclusion
( , , , , , , , , ,PS )F PS GS Y AS U INF R TOT FLIB H EQ 2.7660 0.012 Inconclusive
( , , , , , , , , ,GS )F GS PS Y AS U INF R TOT FLIB H EQ 4.7084 0.000 Cointegration
( , , , , , , , , ,Y )F Y PS GS AS U INF R TOT FLIB H EQ 2.1220 0.047 Inconclusive
( , , , , , , , , ,AS )F AS PS GS Y U INF R TOT FLIB H EQ 2.6908 0.014 Inconclusive
( , , , , , , , , ,U )F U PS GS Y AS INF R TOT FLIB H EQ 3.1875 0.005 Inconclusive
( , , , , , , , , ,R )F R PS GS Y AS U INF TOT FLIB H EQ 2.6692 0.014 Inconclusive
( , , , , , , , , ,INF )F INF PS GS Y AS U R TOT FLIB H EQ 2.3367 0.029 No cointegration
( , , , , , , , , ,TOT )F TOT PS GS Y AS U INF R FLIB H EQ 3.6749 0.002 Cointegration
( , , , , , , , , ,FLIB )F FLIB PS GS Y AS U INF R TOT H EQ 2.7118 0.013 Inconclusive
( , , , , , , , , ,H )F H PS GS Y AS U INF R TOT FLIB EQ 2.3422 0.029 Inconclusive
( , , , , , , , , ,EQ )F EQ PS GS Y AS U INF R TOT FLIB H 4.4042 0.000 Cointegration
Asymptotic critical value bounds are obtained from Table CI(iii), Case V: unrestricted intercept and unrestricted trends for k=10 (Persaran et al: 2001). Lower bound I(0)=2.43 and Upper bound I(1)=3.56 at the 5% significance level. * Denotes significance at the 5% level. ** Denotes significance at the 1% level.
After initially estimating equation (13), the results suggested that social assistance payments as a
proportion of household disposable income (AS), inflation (INF), the real interest rate (R) and the
break-point dummy variable coinciding with the early 1990s recession (B1990) were statistically
insignificant. The following ARDL was estimated:
24
25
i t i− +
−
− +
t
0 11 1 1
p p p
t i t i i t ii i i
PS t PS GS Yα α δ β φ− −= = =
Δ = + + Δ + Δ + Δ∑ ∑ ∑
1 1 1 1
p p p p
i t i i t i i t i i t ii i i i
U TOT FLIB Hγ ρ ψ ξ− − −= = = =
Δ + Δ + Δ + Δ +∑ ∑ ∑ ∑
1 1 2 1 3 1 4 11
p
i t i t t t ti
EQ PS GS Y Uω λ λ λ λ− − − −=
Δ + + + +∑
5 1 6 1 7 1 8 1t t t tTOT FLIB H EQ uλ λ λ λ− − − −+ + + + (15)
The estimated long-run coefficient estimates are provided in Table 13. For the ratio of government
saving to GDP (GS) over the period 1984:1-2006:2, the estimated coefficient is -0.39, and
statistically significant only at the 10 per cent level. For the other variables, a one per cent rise in
household gross disposable income (Y) is estimated to raise the ratio of private savings to GDP by
0.43 per cent. Net foreign liabilities (FLIB) are also significant at the 1 per cent level – and indicate
that Australian financial markets have become more integrated with global capital flows. The
long-run coefficient on the terms of trade (TOT) is slightly higher than the previous estimations,
which possibly indicates that as Australia has become more integrated with the global economy and
that international price determination for traded goods may be exerting a greater influence over
household incomes, consumption and saving. The house price index is now statistically
insignificant, while equity prices remain significant at the 10 per cent level.
Table 13: Estimated long-run coefficients for equation (15) ARDL (2,1,0,2,0,1,0,0) selected lags based on Schwarz Bayesian Criterion
Variable Coefficient Standard Error T-Ratio Probability
Constant -0.3901 0.2294 -1.7001 0.093
Trend -0.0006 0.0004 -1.2942 0.200
GS -0.3855 0.2386 -1.6160 0.110
Y 0.4338 0.2371 1.8295 0.071
U 0.4296 0.3463 1.2407 0.219
TOT 0.0012 0.0003 3.5862** 0.001
FLIB -0.0700 0.0227 -3.0776** 0.003
H 0.0202 0.0242 0.8328 0.408
EQ 0.0341 0.0187 1.8232 0.072 * Denotes significance at the 5% level. ** Denotes significance at the 1% level.
The short-run error correction estimates are presented in Table 14. In the short-run, the error
correction equation indicates a private savings offset of -0.40 to changes in government saving,
which is both statistically significant and roughly equivalent to the estimated long-run coefficient.
The error correction term, , is of the correct sign and statistically significant – indicating
that deviations from the long-run rate of private savings are corrected by around 50 per cent in the
next period.
( 1)ecm −
Table 14: Error correction representation of equation (15) ARDL (2,1,0,2,0,1,0,0) selected lags based on Schwarz Bayesian Criterion
Variable Coefficient Standard Error T-Ratio Probability
Constant -0.1816 0.1008 -1.8006 0.076
Trend -0.0003 0.0002 -1.3609 0.177
( 1)PSΔ − -0.1769 0.0805 -2.1976* 0.031
GSΔ -0.3977 0.1049 -3.7921** 0.000
YΔ 0.2019 0.1110 1.8187 0.073
UΔ -0.4623 0.3714 -1.2445 0.217
( 1)UΔ − -1.1101 0.3230 -3.4367** 0.001
TOTΔ 0.0006 0.0002 3.4544** 0.000
FLIBΔ -0.0776 0.0189 -4.0914* 0.000
HΔ -0.0094 0.0108 0.8707 0.387
EQΔ -0.0158 0.0078 2.0123* 0.048
( 1)ecm − -0.4654 0.0906 -5.1340** 0.000
* Denotes significance at the 5% level. ** Denotes significance at the 1% level. 0.385 * 0.434 * 0.429 * 0.001 * 0.070 * 0.020 * 0.034 *ecm PS GS Y U TOT FLIB H EQ= + − − − + − − +
0.390 * 0.006 * TrendINPT + 2
0.6690R = 2
0.6072R = F-stat [ ](11, 78) 13.7805 0.000F = SER 0.0073=
RSS 0.0041= DW-statistic 2.0543=
Diagnostic statistics for the error correction mechanism (Table 15) are positive, and indicate that the
model is correctly specified.
Table 15: Diagnostic tests on equation (15) LM Test Statistics 2χ statistic Probability
Serial correlation a 2(4)χ 1.8555 0.762
Normality b 2(2)χ 0.4971 0.780
Functional form c 2(1)χ 0.4583 0.498
Heteroscedasticity d 2(1)χ 0.3776 0.539
* Denotes significance at the 5% level. ** Denotes significance at the 1% level. a Breusch-Godfrey LM test for serial correlation. b Jarque-Bera normality test. c Ramsey RESET test for omitted variables/functional form. d White test for heteroscedasticity.
26
27
4. Conclusions
Results from the estimations suggest that while there is no full Ricardian response in Australia to
changes in the fiscal stance, fiscal policy has some ability to impact the real economy. Estimates
suggest a long-run private saving offset around one half, and between -0.25 and -0.40 in the short
run.
While the lower short-run offsets revealed through the error correction mechanisms indicate that
nominal and real frictions and/or rigidities prevent some proportion of the offsetting behaviour
occurring more quickly, this result is consistent with Keynesian models – suggesting that fiscal
policy has a greater ability to influence the real economy over the short term (particularly where
some households are liquidity constrained). While full Ricardian equivalence has not been observed
in the results, they do suggest that over the longer-term, households and organisations are more
forward-looking, and exhibit some partial Ricardian behaviour.
A critical question this paper has also sought to answer is the extent to which the development of
the Australian financial sector (and increased integration into global capital markets) may have
dampened the impact of fiscal policy on the real economy. Estimates of the long-run coefficient on
government saving over the two subsamples (1959:3-1983:4 and 1984:1-2006:2) did not provide
any clear indication that this may be occurring (both sets of estimations produced a long-run
coefficient on government saving around -0.39). However, the short-run error correction
coefficients were markedly different, with the second subsample estimation yielding a short-run
private saving offset that was close to the long-run estimate (-0.40).
Results also confirm greater linkages between Australia and the global economy. While the
coefficient on net foreign liabilities (FLIB), which was taken as a proxy for financial market
openness, was statistically insignificant in the first subsample, this coefficient was found to be
statistically significant in the second subsample. The negative value of this coefficient (-0.07)
suggests that greater access to international capital has lowered private saving. The coefficient on
the terms of trade (TOT) was also higher in the second subsample, which indicates that Australia
may have been deriving higher income from commodities over this period.
While results in this paper suggest that households are not fully Ricardian, fiscal policy can
nonetheless exert some impact on real economic activity. However, it is unreasonable to expect that
any discretionary fiscal policy actions will have a one-for-one impact on the real economy. To the
extent that households anticipate higher (lower) taxes in the future, they will partially offset any
policy action through higher (lower) saving. Where policymakers see a need for discretionary
28
policy, it is important to consider the composition of expenditure, as policies directed at particular
sectors or households will likely generate different impacts.
While there is a role for activist fiscal policy under extreme economic circumstances, the results
also indicate that fiscal policy will only exert a partial impact on activity. It would take substantial
movements in the fiscal stance (greater than 1 per cent of GDP) to have a marked impact on the real
economy. Such large movements in the fiscal position only exacerbate the risks of poor policy,
which includes a risk of excessive debt accumulation, entrenched expenditures and pro-cyclical
impacts (arising from poorly timed policy).
29
References
Aisen, A. & Hauner, D. 2008, Budget Deficits and Interest Rates: A Fresh Perspective, IMF Working Paper No. 08-42, International Monetary Fund, Washington, D.C.
Alesina, A. & Perotti, P. 1995, The political economy of budget deficits, IMF Staff Papers, 42(1),
1-31. Atkins, F. J. 2002, Multiple Structural Breaks in the Nominal Interest Rate and Inflation in Canada
and the United States, Department of Economics Discussion Paper, 2002-07, University of Calgary.
Auerbach, A. J. 2005, The Effectiveness of Fiscal Policy as Stabilisation Policy, mimeo, University
of California, Berkeley. Auerbach, A. J. 2009, Implementing the New Fiscal Policy Activism, Working Paper No. 14725,
National Bureau of Economic Research, Cambridge, Massachusetts. Australian Securities Exchange, 2007, 2006 Australian Share Ownership Study, Australian
Securities Exchange, Sydney. Bai, J. & Perron, P. 2003, Computation and analysis of multiple structural change models, Journal
of Applied Econometrics, 18, 1-22. Bai, J. & Perron, P. 2003, Critical values for multiple structural change tests, Econometrics Journal,
6(1), 72-8. Ball, L. & Mankiw, N, G. 1995, What do budget deficits do?, in, Budget Deficits and Debt: Issues
and Options, A symposium sponsored by the Federal Reserve Bank of Kansas City, Jackson Hole, Wyoming, 5-55.
Banerjee, A. Lumsdaine, R. L. & Stock, J. H. 1992, Recursive and sequential tests of unit-root and
the trend break hypotheses: theory and international evidence, Journal of Business Economics and Statistics, 10, 271-87.
Barro, R, J. 1974, Are government bonds net wealth?, Journal of Political Economy, 82, 1095-
1117. Barro, R, J. 1979, On the determination of the public debt, Journal of Political Economy, 87, 941-
969. Barro, R. J. 1981, Output effects of government purchases, Journal of Political Economy, 89, 1086-
1121. Barro, R, J. 1987, Government spending, interest rates, prices, and budget deficits in the United
Kingdom, 1701-1918, Journal of Monetary Economics, 20, 221-247. Barro, R, J. 1989, The Ricardian approach to budget deficits, Journal of Economic Perspectives,
3(2), 37-54.
30
Bayoumi, T. & Sgherri, S. 2006, Mr Ricardo’s Great Adventure: Estimating Fiscal Multipliers in a Truly Intertemporal Model, International Monetary Fund Working Paper No. 168, International Monetary Fund, Washington D.C.
Ben-David, D. & Papell, D. H. 1995, The great wars, the great crash and the unit root hypothesis, Journal of Monetary Economics, 36, 453-75. Ben-David, D. Lumsdaine, R, L. & Papell, D, H. 2003, Postwar slowdowns and long-run growth:
evidence from two structural breaks, Empirical Economics, 28(2), 303-19. Bernheim, D. 1987, Ricardian Equivalence: an evaluation of theory and evidence, in S. Fischer,
Macroeconomics Annual, pp. 263-304, National Bureau of Economic Research. Bernheim, D. 1989, A Neoclassical perspective on budget deficits, Journal of Economic
Perspectives, 3(2), 55-72. Blanchard, O. & Perotti, R. 2002, An empirical characterization of the dynamic effects of changes
in government spending and taxes on output, The Quarterly Journal of Economics, 117(4), 1329-68.
Byrne, J. P. and Perman, R. 2006, Unit Roots and Structural Breaks: A Survey of the Literature,
Department of Economics Discussion Paper, 2006-10, University of Glasgow. Carrion-I-Silvestre, J. L. & Sanso, A. 2006, Joint hypothesis specification for unit root tests with a
structural break, Econometrics Journal, 9, 196-224. Christiano, L, J. 1992, Searching for a break in GNP, Journal of Business and Economic Statistics,
10(3), 237-50. Clemente, J. Montañés, A. & Reyes, M. 1998, Testing for a unit root in variables with a double
change in the mean, Economics Letters, 59, 175-182. Comley, B. Anthony, S. & Ferguson, B. 2002, The effectiveness of fiscal policy in Australia –
selected issues, Economic Roundup, Winter 2002, 45-72, Canberra. Commonwealth of Australia, 2009, Budget Strategy and Outlook, Budget Paper No.1, 2009-2010,
Australian Government Publishing Service, Canberra. Congressional Budget Office, 2008, Options for Responding to Short-Term Economic Weakness,
January, Washington. D.C. Cotis, J. Coppel, J. & de Mello, L. 2006, Is the United States prone to “overconsumption”?, in, The Macroeconomics of Fiscal Policy, eds, Kopcke, R, W. Tootell, G. & Triest, R, K. MIT Press, Cambridge, Massachusetts. Creel, J. & Sawyer, M. 2009, Current Thinking on Fiscal Policy, Palgrave Macmillan, London. Dawson, J. W. & Strazicich, M. 2006, Time Series Tests of Income Convergence With Two
Structural Breaks: An Update and Extension, Working Papers, Department of Economics, Appalachian State University.
31
Deng, A. & Perron, P. 2006, A Comparison of Alternative Asymptotic Frameworks to Analyse a Structural Change in a Linear Time Trend, Working Paper, Department of Economics, Boston University.
Dickey, D, A. & Fuller, W, A. 1979, Distribution of the estimators for autoregressive time series
with a unit root, Journal of the American Statistical Association, 84(3), 427 31. Dukpa, K. & Perron, P. 2006, Unit Root Tests for a Break in the Trend Function at an Unknown
Time Under Both the Null and Alternative Hypotheses, Working Paper, Department of Economics, Boston University.
Elmendorf, D. W. & Furman, S. 2008, If, When, How: A Primer on Fiscal Stimulus, The Brookings
Institution, Washington, D.C. Elliott, G. Rothenborg, T. J. & Stock, J.H. 1996, Efficient tests for an autoregressive unit root,
Econometrica, 64, 813-36. Engle, R, F. & Granger, C, W, J. 1987, Co-integration and error correction: representation,
estimation and testing, Econometrica, 55(2), 251-76. Engle, R, F. & Yoo, B, S. 1987, Forecasting and testing in cointegrated systems, Journal of
Econometrics, 35, 143-159. Fuller, W. A. 1976, Introduction to Statistical Time Series, John Wiley, New York. Fuller, W. A. 1985, Nonstationary autoregressive time series, in Handbook of Statistics 5: Time
Series in the Time Domain, eds, E, J. Hannan, P, R, Krishnaiah. & M, M, Rao, Elsevier Publishers, Amsterdam.
Gale, W. G. & Orszag, P. R. 2003, Economic effects of sustained budget deficits, National Tax
Journal, 56(3), 463-85. Gale, W. G. & Orszag, P. R. 2004, Budget Deficits, National Saving and Interest Rates, Brookings
Institution, Washington, D. C. Garcia, R. & Perron, P. 1996, An analysis of the real interest rate under regime shifts, The Review of
Economics and Statistics, 78,111-25. Glynn, J. Perera, N. & Verma, R. 2007, Unit root tests and structural breaks: A survey with
applications, Journal of Quantitative Methods for Economics and Business Administration, 3(1), 63-79.
Granger, C, W, J. & Newbold, P. 1974, Spurious regressions in econometrics, Journal of
Econometrics, 2, 111-20. Gregory, A, W. & Hansen, B. 1996, Residual based tests for cointegration in models with regime
shifts, Journal of Econometrics, 70, 199-226. Gregory Mankiw, N. 2000, The savers-spenders theory of fiscal policy, Papers and Proceedings of
the One Hundred and Twelfth Annual Meeting of the American Economic Association, 90(2), edS, Baldwin, D, & Oaxaca, R, L.
32
Gruen, D. & Sayegh, A. 2005, The evolution of fiscal policy in Australia, Oxford Review of Economic Policy, 21(4), 618-35.
Hansen, B, E. 2001, The new econometrics of structural change: Dating breaks in US labor productivity, Journal of Economic Perspectives, 15, 117-28. Harvey, D, I. & Mills, T, C. 2004, Tests for stationarity in series with endogenously determined
structural change, Oxford Bulletin of Economics and Statistics, 66, 863-94. Hauner, D. 2006, Fiscal Policy and Financial Development, IMF Working Paper, No. 06-26,
International Monetary Fund, Washington, D.C. Hauner, D. & Kumar, M.S. 2006, Fiscal Policy and Interest Rates: How Sustainable is the New
Economy?, IMF Working Paper No. 06-112, International Monetary Fund, Washington, D.C. Hemming, R. Kell, M. & Mahfouz, S. 2002, The Effectiveness of Fiscal Policy in Stimulating
Economic Activity – a Review of the Literature, International Monetary Fund Working Paper No. 02-208, International Monetary Fund, Washington, D.C.
International Monetary Fund, 2008, World Economic Outlook: Fiscal Policy as a Countercyclical
Tool, October, International Monetary Fund, Washington D.C. Johansen, S. 1991, Estimation and hypothesis testing of cointegration vector in Gaussian vector
autoregressive models, Econometrica, 59(6), 1551-80. Johansen, S. 1995, Likelihood-Based Inference in Cointegrated Vector Autoregressive Models,
Oxford University Press. Johansen, S. & Juselius, K. 1990, Maximum likelihood estimation and inference on cointegration –
with applications to the demand for money, Oxford Bulletin of Economics and Statistics, 52(2), 169-210.
Kennedy, S. Luu, N. Morling, S. & Yeaman, L. 2004, Fiscal Policy in Australia: Discretionary
Policy and Automatic Stabilisers, paper prepared for the Treasury/Australian National University Macroeconomic Conference, Canberra.
Kim, K, D. & Perron, P. 2008, Assessing the Relative Power of Structural Break Tests Using a
Framework Based on the Approximate Bahadur Slope, Working Paper, Boston University. Kim, K, D. & Perron, P. 2009, Unit root tests allowing for a break in the trend function at an
unknown time under both the null and alternative hypotheses, Journal of Econometrics, 148(1), 1-13.
Kirchner, S. 2007, Fiscal policy and interest rates in Australia, Policy, 23(3), 11-15. Kopcke, R. W. Tootell, G, M, B. & Triest, R. K. 2006, The Macroeconomics of Fiscal Policy, MIT
Press, Cambridge, Massachusetts. Kremers, J, J. Ericson, N, R. and Dolado, J, J. 1992, The power of cointegration tests, Oxford
Bulletin of Economics and Statistics, 54, 545-79.
33
Krugman, P. 2005, Is fiscal policy poised for a comeback?, Oxford Review of Economic Policy, 21(4), 515-23.
Leachman, L, L. 1996, New evidence on the Ricardian equivalence theorem: a multicointegration
approach, Applied Economics, 28, 695-704. Lee, J. Huang, C, J. & Shin, Y. 1997, On stationary tests in the presence of structural breaks,
Economics Letters, 55, 165-72. Lee, J. 1999, Stationarity tests with multiple endogenized breaks, In Rothman, P. eds, Nonlinear
Time Series Analysis of Economic and Financial Data, Kluwer Academic Press, Massachusetts.
Lee, J. & Strazicich, M. 2001, Testing the null hypothesis of stationarity in the presence of
structural breaks, Applied Economic Letters, 8, 377-82. Lee, J. & Strazicich, M. 2003, Minimum Lagrange multiplier unit root tests with two structural
breaks, Review of Economics and Statistics, 81, 1082-1089. Lee, J. & Strazicich, M. 2004, Minimum LM Unit Root Test With One Structural Break, Working
Papers, No. 04-17, Department of Economics, Appalachian State University. Lee, J. List, A. J. & Strazicich, M. 2005, Modelling Nonrenewable Resource Prices: Deterministic
or Stochastic Trends?, NBER Working Paper Series, No. 11487, National Bureau of Economic Research, Cambridge, Massachusetts.
Leiderman, L. & Bleijer, M, I. 1988, Modelling and testing Ricardian equivalence: a survey, IMF
Staff Papers, 35, 1-31. Leybourne, S, J. & Newbold, P. 2003, Spurious rejections by cointegration tests induced by
structural breaks, Applied Economics, 35(9), 1117-21. Liu, H. & Rodriguez, G. 2006, Unit root tests and structural change when the initial observation is
drawn from its unconditional distribution, Econometrics Journal, 9, 225-51. Lopez, C. 2005, Improved Unit Root Tests With Changes in the Intercept, Working Paper,
Department of Economics, University of Cincinnati. Lopez, C. Murray, C, J. & Papell, D. 2005, State of the art unit root tests and purchasing power
parity, Journal of Money, Credit, and Banking, 37, 361-69. Lumsdaine, R, L. & Papell, D, H. 1997, Multiple trend breaks and the unit root hypothesis, Review
of Economics and Statistics, 79(2), 212-18. Makin, T. 2007, Re-examining the effectiveness of stabilisation policy, Australian Economic
Papers, 46(4), 348-59. De Mello, L. Kongsrud, P, M. & Price, R. 2004, Saving Behaviour and the Effectiveness of Fiscal
Policy, Economics Department Working Papers, No. 397, Organisation for Economic Co-Operation and Development, Paris.
34
Narayan, P, K. and Smyth, R. 2004, Structural breaks and unit roots in Australian macroeconomic time series, Department of Economics Discussion Paper, No. 18/04, Department of Economics, Monash University.
Nelson, C. & Plosser, P. 1982, Trends and random walks in macroeconomic time series: some
evidence and implications, Journal of Monetary Economics, 10, 139-62. Nunes, L, C. C-M, Kuan & Newbold, P. 1997, Spurious regression, Econometric Theory, 11, 736-
49. Nunes, L, C. Newbold, P. & C-M, Kuan. 1997, Testing for unit roots with breaks: Evidence on the great crash and the unit root hypothesis reconsidered, Oxford Bulletin of
Economics and Statistics, 59, 435-48. Ohara, H, I. 1999, A unit root test with multiple trend breaks: A theory and application to US and
Japanese macroeconomic time series, The Japanese Economic Review, 50, 266-290. Organisation for Economic Co-operation and Development (OECD), 2009, OECD Interim
Economic Outlook Chapter 3: The Effectiveness and Scope of Fiscal Stimulus, OECD, Paris. Papell, D, H. 1997, Searching for Stationarity: Purchasing power parity under the current float,
Journal of International Economics, 43, 313-332. Papell, D, H. & Prodan, R. 2003, The uncertain unit root in US real GDP: Evidence with restricted and unrestricted structural change, Journal of Money, Credit, and Banking, 36,
423-427. Papell, D, H. & Prodan, R. 2006, Additional evidence of long-run purchasing power parity with
restricted structural change, Journal of Money, Credit, and Banking, 38, 1329-49. Perotti, R. 2002, Estimating the effects of fiscal policy in OECD countries, European Central Bank
Working Paper No. 168. Perotti, R. 2005, Estimating the effects of fiscal policy in OECD countries, Discussion Paper No.
4842, Centre for Economic Policy Research, London. Perotti, R. 2007, In search of the transmission mechanism of fiscal policy, Working Paper No.
13143, National Bureau of Economic Research, Cambridge, Massachusetts. Perron, P. 1989, The great crash, the oil price shock, and the unit root hypothesis, Econometrica,
57(6), 1361-401. Perron, P. 2006, Dealing with structural breaks, in Palgrave Handbook of Economics, Vol. 1:
Econometric Theory, eds, Patterson, K. and Mills, T, C., Palgrave Macmillan. Perron, P. & Vogelsang, T. 1992, Nonstationarity and level shifts with an application to purchasing
power parity, Journal of Business and Economic Statistics, 10, 301-20. Perron, P. 1994, Unit root and structural change in macroeconomic time series, in Cointegration for
the Applied Economist, eds, B. Rao, Macmillan, London.
35
Perron, P. 1997, Further evidence on breaking trend functions in macroeconomic variables, Journal of Econometrics, 80(2), 355-85.
Perron, P. 2006, Dealing with structural breaks, Palgrave Handbook of Econometrics, Volume 1:
Econometric Theory, eds, T, C, Mills. & Patterson, K. Palgrave Macmillian, London. Pesaran, M, H. Shin, Y. & Smith, R, J. 1996, Testing for the Existence of a Long-run Relationship,
Cambridge Working Papers in Economics, Department of Applied Economics, Cambridge University.
Pesaran, M, H. 1997, The role of economic theory in modelling the long run, Economic Journal,
107(440), 178-91. Pesaran, M, H. & Shin, Y. 1998, An autoregressive distributed-lag modelling approach to
cointegration analysis, Econometrics and Economic Theory in the Twentieth Century: The Ragnar Frisch Centennial Symposium, eds, Strom, S.
Pesaran, M, H. & Smith, R, J. 1998, Structural analysis of cointegrating VARs, Journal of
Economic Surveys, 12(5), 471-505. Pesaran, M, H. Shin, Y. & Smith, R, J. 2001, Bounds testing approaches to the analysis of level
relationships, Journal of Applied Econometrics, 16(3), 289-326. Ricardo, D. 1966, The Works and Correspondence of David Ricardo, 1, edited by Piero Sraffa,
Cambridge University Press, London. Ricardo, D. 1966, The Works and Correspondence of David Ricardo, 4, edited by Piero Sraffa,
Cambridge University Press, London. Ricciuti, R. 2003, Assessing Ricardian Equivalence, Journal of Economic Surveys, 17(1), 57-78. Seater, J, J. 1993, Ricardian equivalence, Journal of Economic Literature, 31, 142-190. De Serres, A. Pelgrin, F. 2002, The decline in private savings rates in the 1990s in OECD
countries: How much can be explained by non-wealth determinants?, Working Paper No. 344, Organisation for Economic Cooperation and Development, Paris.
Sheffrin, S, M. & Woo, W, T. 1990b, Testing an optimizing model of the current account via the
consumption function, Journal of International Money and Finance, 9, 220-233. Smyth, R. & Kumar Narayan, P. 2005, Structural breaks and unit roots in Australian
macroeconomic time series, Pacific Economic Review, 10, 421-37. Solow, R, M. 2005, Rethinking fiscal policy, Oxford Review of Economics and Statistics, 21(4),
509-14. Spilimbergo, A. Symansky, S. Blanchard, O. & Cottarelli, C. 2008, Fiscal policy for the crisis, IMF
Staff Position Note 08/01, International Monetary Fund, Washington, D. C. Spilimbergo, A. Symansky, S. & Schindler, M. 2009, Fiscal multipliers, IMF Staff Position Note
09/11, International Monetary Fund, Washington, D. C.
36
Strazicich, M. Lee, J. & Day, E. 2004, Are incomes converging among OECD countries? Time series evidence with two structural breaks, Journal of Macroeconomics, 26, 131-145.
Taylor, B. 2000, Reassessing discretionary fiscal policy, Journal of Economic Perspectives, 14(3),
21-36. Westerlund, J. 2006, Testing for panel cointegration with multiple structural breaks, Oxford Bulletin
of Economics and Statistics, 68, 101-32. Westerlund, J. 2006, Simple Unit Root Tests With Multiple Breaks, Working Paper, Department of
Economics, Lund University. Yule, G, U. 1926, Why do we sometimes get nonsense correlations between time series? A study in
sampling and the nature of time series, Journal of the Royal Statistical Society, 89, 1-64. Zivot, E. & Andrews, D. 1992, Further evidence on the great crash, the oil price shock, and the unit
root hypothesis, Journal of Business and Economic Statistics, 10(3), 251-70.